Phylogeny and Biogeography of Tribe Hibisceae (Malvaceae) on Madagascar
Margaret M. Koopman and David A. Baum
Department of Botany, University of Wisconsin, Madison, Wisconsin 53706 U.S.A.
Author for correspondence (email@example.com)
Communicating Editor: Gregory M. Plunkett
Abstract—Madagascar boasts a substantial radiation within Hibisceae (Malvaceae). Molecular data were used to determine the number of
migrations to Madagascar, to clarify the timing of these radiations, and to assess the relationships among the endemic taxa. Sequences of matK
and ndhF were obtained for representatives of all the Malagasy endemic genera in Hibisceae, as well as representative Malagasy species of
Hibiscus and Kosteletzkya. Phylogenetic analysis of these data along with a sampling of outgroups and other major lineages of Hibisceae
strongly supported the existence of an exclusively Malagasy clade, the /Megistohibiscus clade, which is sister to the rest of tribe Hibisceae.
Humbertiella,Megistostegium,Perrierophytum, all the Malagasy Kosteletzkya, and a number of endemic species usually placed in Hibiscus are
members of /Megistohibiscus and appear to be derived from a single introduction to Madagascar in the mid Miocene. Other sampled
Malagasy Hibisceae, including the endemic genera Helicteropsis,Jumelleanthus, and Macrostelia, were resolved as members of the /Euhibiscus
clade, implying at least one additional dispersal to the island.
Keywords—biogeography, Hibisceae, Madagascar, Malvaceae, matK/ndhF, molecular dating.
Madagascar is celebrated for its diverse and highly en-
demic flora. The diversification of Hibisceae (Malvaceae;
Malvoideae) on Madagascar provides an opportunity to elu-
cidate the mechanisms responsible for the island’s great bio-
logical diversity, which remains poorly understood (Wilmé
et al. 2006). Madagascar boasts at least 86 species of Hibisceae
(Table 1), of which approximately 75% are endemic (Ho-
chreutiner 1955; Table 1). The island harbors ca. 47 species
traditionally assigned to Hibiscus L., of which 32 (68%) are
endemic (Hochreutiner 1955). Macrostelia Hochr., a genus
once thought to be endemic to Madagascar, was expanded by
the addition of a few Australian species (Hochreutiner 1952;
Fryxell 1974a), but Craven and Pfeil (2004) subsequently
transferred the Australian taxa to Hibiscus. Thus, currently,
Madagascar is home to six endemic genera, generally as-
signed to Hibisceae (Hochreutiner 1955; Dorr 1990; Bayer and
Kubitzki 2003): Helicteropsis Hochr. (one species), Humber-
tiella Hochr. (six species), Jumelleanthus Hochr. (one species),
Macrostelia (three species), Megistostegium Hochr. (three spe-
cies), and Perrierophytum Hochr. (nine species). A seventh
endemic genus, Humbertianthus Hochr., includes one species
(H. cardiostegius Hochr.) that is synonymous with Macrostelia
laurina Hochr. & Humbert (M. M. Koopman, unpublished
The historical recognition of six or more endemic genera of
Hibisceae in Madagascar reflects the existence of consider-
able morphological diversity within the group. The genus
Humbertiella has unusual and elaborate staminal glands
called coronules (corresponding to inflated, apical teeth; Dorr
1990). Megistostegium flowers have an enormous red cam-
panulate epicalyx. Perrierophytum species range from small,
sclerophyllous shrubs with inconspicuous flowers to species
with showy inflorescences of white, violet, or green flowers.
Helicteropsis is characterized by unusual leaves with a cordate
base and deeply emarginated apex, reduced petals, a red
tubular calyx, an elongated staminal tube, and a four-locular
ovary. The monotypic genus Jumelleanthus is unique in Hi-
bisceae in having two collateral ovules per locule (Bayer and
A comprehensive understanding of the evolutionary
mechanisms responsible for the radiation of Hibisceae on
Madagascar is contingent on a well resolved phylogeny for
the group. Additionally, a phylogeny will improve our un-
derstanding of the number and timing of independent inva-
sions into Madagascar. In this study, sequence data from the
plastid genes ndhF and matK were used to examine the phy-
logenetic and biogeographical relationships among the Mala-
gasy Hibisceae, and their relationships to other elements of
the group worldwide.
/Eumalvoideae (Baum et al. 2004), the clade formerly rec-
ognized as the mallow family, Malvaceae s.s., was tradition-
ally divided into five tribes: Decaschistieae, Gossypieae, Hi-
bisceae, Malvavisceae, and Malveae. Subsequent molecular
phylogenetic analysis showed that Decaschistieae and Mal-
vavisceae are embedded within a paraphyletic Hibisceae (in-
deed within a paraphyletic genus Hibiscus), resulting in the
proposal to merge these three taxa into a single broadly con-
strued Hibisceae (Pfeil et al. 2002). Bayer and Kubitzki (2003)
followed the proposal to expand Hibisceae, but they also
recognized an additional tribe, Kydieae, to accommodate
four Asian genera of unknown affinities. Subsequently, Ky-
dieae was subsumed within Hibisceae based on evidence
that, like Decaschistieae and Malvavisceae, its component
species fall within Hibisceae s.l. (Pfeil and Crisp 2005).
Molecular phylogenetic analysis and morphological data
have shown that Gossypieae, Malveae, and Hibisceae s.l. are
all monophyletic (Seelanan et al. 1997; Pfeil et al. 2002; Tate et
al. 2005). Additionally, Gossypieae, Malveae, and the Aus-
tralian genus Alyogyne Alefeld form a clade that is sister to
Hibisceae (Baum et al. 2004). No synapomorphies exist for
TABLE 1. Genera of tribe Hibisceae present on Madagascar, the num-
ber of species they contain (based on Hochreutiner 1955), the number
endemic to Madagascar, and their geographic distribution.
(Endemic) Distribution and habitat on Madagascar
Abelmoschus 3 (0) Widespread
Hibiscus s.l. 47 (32) Widespread
Helicteropsis 1 (1) N, dry deciduous forests on karstic limestone
Humbertiella 6 (6) SW, xeric scrubland/spiny forest
Jumelleanthus 1 (1) N, wet forests
Megistostegium 3 (3) SW, xeric scrubland/spiny forest
Perrierophytum 9 (9) W, xeric scrubland/dry deciduous forest
Kosteletzkya 9 (8) SW, xeric scrubland; Central, high plateau
Macrostelia 3 (3) E, wet littoral forests
Pavonia 3 (1) Central, high plateau
Urena 1 (0) Widespread
Total 86 (66)
Systematic Botany (2008), 33(2): pp. 364–374
© Copyright 2008 by the American Society of Plant Taxonomists
Hibisceae, although the plesiomorphic presence of apical
teeth on the staminal column (shared with Gossypieae) and
branching styles (shared with Malveae) serves to distinguish
species of Hibisceae from Malveae or Gossypieae (Pfeil et al.
Current generic classification within Hibisceae does not
adequately reflect evolutionary relationships because Hibis-
cus, as traditionally understood, is massively paraphyletic
(Pfeil et al. 2002). Among the taxa embedded within Hibiscus
are segregate genera whose affinities with Hibiscus were al-
ways appreciated (e.g. Talipariti Fryxell, Fioria Mattei, Ceno-
centrum Gagnep.) as well as more distinctive taxa, such as
Urena L. and Pavonia Cav. In response to the phylogenetic
results, Pfeil and Crisp (2005) suggested a broad circumscrip-
tion of Hibiscus to encompass the smallest clade that includes
all Hibiscus species, with the exception of Malagasy members
of Hibiscus section Azanza DC. whose taxonomic status was
Gossypieae, Malveae, Alyogyne, and Hibisceae together
form the clade /Eumalvoideae, which is embedded in a
larger clade, /Malvoideae. Relative to /Eumalvoideae, the
remainder of Malvoideae form a grade of smaller genera
previously assigned to Bombacaceae (Quararibea Aubl., Ma-
tisia Bonpl., Phragmotheca Cuatrec., Pentaplaris L. O. Williams
and Standl.) or Malvaceae s.s. (Uladendron Marcano-Berti,
Radyera Bullock, Howittia F. Muell, Lagunaria G. Don, Camp-
tostemon Mast.). The earlier branches of this grade (former
Bombacaceae plus Uladendron) are Neotropical rainforest
trees, whereas Radyera,Howittia,Lagunaria, and Camptostemon
show an Asian or Australasian distribution (Radyera also oc-
curs in South Africa). Consequently, it has been suggested
that Malvoideae underwent transPacific dispersal from the
Neotropics, with subsequent worldwide dispersal of /Eum-
alvoideae from an Australasian cradle (Pfeil et al. 2002; Baum
et al. 2004). These dispersal events (Malvoideae out of the
New World and the /Eumalvoideae radiation) were accom-
panied by an accelerated rate of molecular evolution in the
plastid genome (Baum et al. 2004). Given this biogeographic
hypothesis, the large diversity of Hibisceae present on Mada-
gascar is noteworthy, especially if the invasion of Madagas-
car occurred during or soon after the radiation of /Eumal-
voideae. Here we show that at least two separate migrations
of Hibisceae to Madagascar occurred, each of which resulted
in significant taxonomic and morphological diversification.
One of these gave rise to a clade that is sister to the rest of
Hibisceae and is likely to have diverged early in the radiation
MATERIALS AND METHODS
Sampling—We obtained 31 ndhF and 30 matK sequences, which were
combined with previously published sequences to create a complete data
set for 50 taxa (Appendix 1; TreeBASE study number S1846). Ingroup
sampling included all of the Malagasy endemic genera, two of the eight
endemic Kosteletzkya species and five Hibiscus species endemic to the
island. The Australian Hibiscus tozerensis Craven and B. E. Pfeil (previ-
ously Macrostelia grandifolia Fryxell) was included, as were 10 additional
species of non-Malagasy Hibiscus that were selected to provide a good
representation of biogeographical areas and well-demarcated clades
(guided by Pfeil et al. 2002; Pfeil and Crisp 2005). Sequences from the
enigmatic genus Kydia Roxb. were also obtained. Previously published
sequences for representatives of Malveae and Gossypieae were included
(Baum et al. 2004), as were sequences for the Afro-Malagasy taxon Gos-
sypioides kirkii (Mast.) Skovsted (Seelanan et al. 1997). Outgroup taxa in-
cluded the genera Howittia,Radyera,Camptostemon,Lagunaria, and the
monotypic Neotropical tree genus, Uladendron (Baum et al. 2004). For ease
of communication, taxa will be referred to by their generic names, except
in cases where more than one species was sampled from a given genus
(see Appendix 1).
DNA Extraction and Sequencing—Total genomic DNA was extracted
from fresh, silica-dried or herbarium leaf tissue as described in Alverson
et al. (1999). The ndhF gene was amplified as overlapping fragments as
described in Baum et al. (2004). The matK region, comprising the trnK
intron and matK coding sequence, was amplified using primers as de-
scribed in Nyffeler et al. (2005). PCR products were purified using AM-
Pure beads (Agencourt Bioscience Corp., Beverley, Massachusetts) and
cycle-sequenced (Big Dye v. 3.1, Applied Biosystems, Foster City, Cali-
fornia) using the manufacturers’protocols. All PCR products were se-
quenced in both directions. Sequences were edited and assembled in
Sequencher 4.1 (Gene Code Corp., Ann Arbor, Michigan), imported into
MacClade (Maddison and Maddison 2003), and aligned manually.
Phylogenetic Analyses—The incongruence length difference (ILD) test
(Farris et al. 1994), implemented as the partition homogeneity test in
PAUP*4.0b10 (Swofford 2001) using simple taxon addition, TBR searches,
holding ten trees at each step, and with maxtrees set to 100, was used to
estimate incongruence between the ndhF and matK data sets. Exploration
of sources of conflict entailed repeating ILD tests with selected taxa
Maximum parsimony (MP) analyses were performed in PAUP*4.0b10.
MP heuristic searches used 100 random taxon addition replicates (hold-
ing one tree at each step) and TBR branch swapping. All characters were
equally weighted, and gaps were treated as missing data. To estimate
clade support, we conducted parsimony bootstrap (BS) analysis using
10,000 bootstrap replicates with simple taxon addition (holding one tree
at each step) and TBR branch swapping. Templeton tests (Templeton
1983), implemented as the Wilcoxon sign-rank tests in PAUP*4.0b10,
were used to explore alternative topologies in a MP framework.
Models of molecular evolution were evaluated using MrModeltest v.
2.0 (Nylander 2004) in the context of trees derived from MP analysis. The
model identified as optimal by hierarchical likelihood ratio tests was used
for maximum likelihood (ML) and Bayesian MCMC analyses.
ML trees were estimated by asis addition sequence, TBR heuristic
searches, with likelihood parameters fixed after optimization on the MP
Bayesian MCMC analysis was implemented with MrBayes v. 3.0
(Huelsenbeck and Ronquist 2001). Three independent MCMC runs were
conducted, each composed of four linked chains (sequential heat = 0.2)
that ran for 5,000,000 generations with sampling every 100 generations.
The burn-in period was estimated by visual examination of a likelihood-
by-generation plot. To evaluate mixing, acceptance values were examined
and the three majority rule consensus trees were compared. Once ad-
equate mixing was achieved, the three posterior distributions were
pooled to obtain the best estimates of clade posterior probabilities (PP).
Because MrBayes places a zero prior probability on a polytomy, it can
yield inappropriately high PPs for short internal branches (Lewis et al.
2005). To detect branches that might be subject to this phenomenon, we
conducted likelihood ratio tests to compare the fully resolved ML tree
versus trees in which each branch in turn was collapsed to a polytomy (as
described in Baum et al. 2004).
Fossil Calibration—A small number of Malvaceaeous fossil pollen
grains can be used as calibration points. The earliest Malvoideae-like
fossil pollen known, Echiperiporites estelae, was found in Venezuela and is
dated to the late Eocene [ca. 37–40 million years ago (Ma); Germeraad et
al. 1968; Muller 1981]. The second fossil pollen record, Malvacearumpollis
bakonyensis, is known first from the Murray Basin of Australia and marine
sediments in southeast Queensland from the late Eocene/early Oligocene
boundary (ca. 35–37 Ma; Wood 1986; MacPhail and Truswell 1989;
MacPhail 1999). No Malvaceae fossil pollen is known from Madagascar.
To properly place these fossil pollen grains on the estimated phylogeny,
palynological traits were compared between the fossils and living species.
Neotropical taxa that are well supported as early-branching members of
Malvoideae (Baum et al. 2004) have pollen that lack echinae altogether
(Matisieae, Nilsson and Robyns 1986) or have minute (∼1m), widely
dispersed echinae (Pentaplaris, Bayer and Dorr 1999), whereas Uladendron
(Marcano-Berti 1971) and the remaining malvoids have larger echinae.
Camptostemon,Howittia,Lagunaria, and members of Malveae and Gos-
sypieae have short, often curved echinae that taper to an acuminate tip
and sit upon a bulbous base. With the exception of the well-nested genus
Lavatera, pollen grains in these groups do not exceed approximately
80m (el Naggar 2004). The tribe Hibisceae, in contrast, have large (88–
130m) pollen grains with long, slender, straight echinae that lack a
bulbous base. Also, as noted by Christensen (1986), Hibisceae pollen
2008] KOOPMAN & BAUM: MALAGASY HIBISCEAE 365
generally has more apertures (from 22 to well over 100) than the other
tribes in /Eumalvoideae.
Grains of the fossil pollen E. estelae are 55–87m in diameter, have
20–24 pores, conical echinae (6–9m) that are 9–14m apart and have a
bulbous base, being thickest at half their height (Germeraad et al. 1968).
The pollen of M. bakonyensis is larger (85–105m), but only 4 pores are
visible. The tall acute-tipped echinae (6–12m tall) are found upon a
bulging base cushion and are 6–10m apart (Wood 1986; MacPhail and
Because these fossil pollen grains have features common among /Eum-
alvoideae, but lack diagnostic characteristics of any of the major sub-
clades, it is not easy to assign them to a specific phylogenetic position
based on morphology. We used the Neotropical E. estelae pollen to fix the
minimal age of the most common recent ancestor of the Neotropical
Uladendron and the ancestrally Australian /Eumalvoideae (Baum et al.
2004). We recognize that even accepting the biogeographic hypothesis of
a single migration from South America to Australasia, this pollen form
could attach deeper in the Neotropics. However, we suggest that this
assignment will provide an approximate first estimate of the time frame
of the Eumalvoid radiation. The Australasian M. bakonyensis pollen could
not be assigned specifically to any of the nested clades within /Eumal-
voideae. We used M. bakonyensis to constrain the minimal age of the last
common ancestor of Radyera (the earliest diverging Australasian taxon;
Baum et al. 2004) and the remainder of /Eumalvoideae. The range of
dates associated with the stratigraphic assignments above (Echiperiporites
estelae,37–40 Ma; Malvacearumpollis bakonyensis,35–37 Ma) were used for
molecular dating analyses.
Molecular Dating—We used a likelihood ratio test to determine wheth-
er a uniform molecular clock could be rejected for these data (Baldwin
and Sanderson 1998). Then, to accommodate branch-to-branch rate het-
erogeneity, ML branch lengths, as estimated without a clock, were rate-
smoothed using Penalized Likelihood (PL, Sanderson 2002), as imple-
mented in r8S v. 1.70 (Sanderson 2003). We used a Truncated-Newton
algorithm with cross-validation to find the optimal smoothing parameter.
Confidence intervals were estimated with a three-step parametric boot-
strap procedure (based on Baldwin and Sanderson 1998): (i) 100 pseudo-
replicate data matrices were generated in the program SeqGen v. 1.3.2
(Rambaut and Grassly 1997) by evolving sequences randomly according
to the maximum likelihood tree and model; (ii) the matrices were im-
ported into PAUP* 4.0 and branch lengths were obtained on the ML tree
topology for each data matrix using newly estimated maximum likeli-
hood parameters; (iii) trees were converted into chronograms using PL in
r8s. The distance (proportional to time) between a particular tip (or col-
lection of tips, i.e. a clade) and an internal node of interest was measured
and collated across all the trees using the “ape”package (Paradis et al.
2004, 2006) in R (R Development Core Team 2005) to obtain a distribution
of divergence times for each node.
Phylogenetic Analyses of Individual Data Sets—The ndhF
and matK data sets have aligned lengths of 2130 and 2822
bases, with 201 and 274 parsimony-informative sites, respec-
tively. Parsimony analysis of the two data sets yielded con-
sensus trees with many well-supported nodes (as judged by
BS and PP; Fig. 1). MrModeltest suggested that the GTR + I
+ G model of molecular evolution was preferred for both
ndhF and matK. For each gene, ML searches yielded trees
with topologies very similar to those obtained by MP (data
Three Bayesian MCMC runs for each data set resulted in
sets of 50,001 trees. The first 10,001 trees were deleted as
burn-in (based on a conservative evaluation of a likelihood
by generation plot) and the remaining 40,000 trees were used
to generate a 50% majority rule consensus tree. For both
data sets, the three runs yielded the same 50% majority rule
consensus tree, which closely resembled the MP trees (not
shown). Clade posterior probabilities were very similar be-
tween the three runs, never differing by more than 3%. Com-
bined with acceptance rates, which were within a suitable
range, we inferred that good mixing occurred during the
MCMC analysis. The PP of clades found in the consensus
trees are shown in Fig. 1.
Evaluation of Discordance Between the Data Sets—An ILD
test shows the existence of significant conflict between ndhF
and matK for the entire data set (p = 0.001). Visual examina-
tion showed that the two data sets yielded very similar to-
pologies for a majority of the taxa but differed significantly in
the placement of three taxa: Helicteropsis, Hibiscus grandidieri,
and Jumelleanthus. These three “problem”taxa were located
at the base of Hibisceae or /Eumalvoideae in the ndhF tree,
whereas the matK data suggested that they are all members of
the /Euhibiscus clade of Hibiscus s.l. (as defined in Table 2).
Three single basepair indels in matK supported the matK to-
pology (see below). If the problem taxa are forced into the
/Euhibiscus clade using a constrained MP search for the
ndhF data, the optimal tree is significantly longer than the
unconstrained ndhF tree (Templeton test; cost = 41 steps; p<
0.0001). Likewise, the matK data reject a clade that includes
all Hibisceae except the three problem taxa (cost = 234 steps;
p < 0.0001). Deletion of the three problem taxa resulted in a
non-significant (p = 0.07) ILD test result, implying that these
three taxa are, indeed, the only major source of genealogical
conflict between the two data sets. Neither the whole se-
quences nor unassembled sequence traces of the problem
taxa were identical to other taxa in the data matrix. Therefore,
we conclude that the alternative placement of these three taxa
is not due to contamination but instead is due to a phyloge-
netic artifact. One possibility worth considering is that the
ndhF genes in these species have been evolving anomalously,
which could have resulted in them being subject to long-
branch attraction or a similar problem. Based on morphologi-
cal data, the position of Helicteropsis, Hibiscus grandidieri and
Jumelleanthus suggested by matK is more plausible (see Dis-
cussion) and is also the result obtained from combined analy-
sis of the two data sets (not shown). Additionally, our un-
published ITS data supports the matK placement of He-
licteropsis with high support. In light of these findings we
opted to delete Helicteropsis, Hibiscus grandidieri and Jumel-
leanthus from the combined analysis.
Combined Phylogenetic Analyses—Parsimony analyses of
the combined data set (minus Helicteropsis, Hibiscus grandi-
dieri, and Jumelleanthus) resulted in 12 MP trees (Fig. 2). Like-
lihood ratio tests in MrModeltest indicated that GTR+I+G
was the best fitting model. Maximum likelihood (ML) analy-
ses produced a topology that was very similar to that ob-
tained by MP (Fig. 3).
All three Bayesian MCMC runs yielded the same 50% ma-
jority rule consensus tree. Clade posterior probabilities were
very similar between the runs, never differing by more than
2%. Combined with acceptance probabilities that fell in an
acceptable range, our results suggest good mixing and, there-
fore, we combined the post burn-in trees into a single pool of
120,000 that was used as a basis for the calculation of the
posterior probabilities for clades on the ML tree (Fig. 3). Five
internal branches, marked with arrows in Fig. 3, were not
significantly different from zero length based on a likelihood
The overall structure of the tree estimated with the com-
bined data (Fig. 2) is fully consistent with the summary tree
provided by Pfeil and Crisp (2005). Hibisceae is divided into
five major clades: /Phylloglandula (which includes the sub-
clades /Furcaria and /Azanza), /Trionum, /Calyphylli,
/Euhibiscus, and a Malagasy clade (which includes Humber-
366 SYSTEMATIC BOTANY [Volume 33
tiella,Megistostegium,Perrierophytum, and Malagasy Kosteletz-
kya). The latter clade was not assigned a name by Pfeil and
Crisp (2005) so we will here refer to this clade as /Megisto-
hibiscus, a reference to the inclusion of one of the most dis-
tinctive segregate genera, Megistostegium. We provide a
branch-based definition of /Megistohibiscus in Table 2.
Structural Molecular Characters—We examined the data
for parsimony-informative indel characters. Fourteen indels
were found in matK and5inndhF that were non-
homoplasious on the ML tree (Fig. 2). One deletion (C) sup-
ports Hibiscus sensu Pfeil et al. (2002) and is missing from
/Megistohibiscus. One insertion (D) supports the placement
of Asian members of Hibiscus section Azanza [Hibiscus mac-
rophyllus Roxb., H. hamabo Siebold and Zucc., H. tiliaceus (syn-
onymous with Talipariti tiliaceum)], as well as Kydia in the
/Phylloglandula clade (sensu Pfeil et al. 2002). No indels
support the /Euhibiscus clade (sensu Pfeil et al. 2002) in its
entirety, although a 1-bp insertion (f)inmatK does support
the smallest clade that includes all the putative /Euhibiscus
taxa sampled from Madagascar. This insertion is also present
in Helicteropsis and Hibiscus grandidieri sequences and thus
supports the matK placement of these taxa (data are missing
in this region for Jumelleanthus). Two indels (kand B) support
the biogeographically disjunct clade of the Asian H. rosa-
sinensis and the Hawaiian H. waimeae. Within /Megistohibis-
cus, the monophyly of both Humbertiella and Megistostegium
is well-supported, and each clade is demarcated by two in-
dels (Fig. 2).
One remarkable structural character is an inversion in
matK comprising a core 5-bp inversion (GGGAA or TTCCC)
bordered on both sides by a perfect 16-bp inverted repeat.
The TTCCC version of the core sequence is found in all mem-
bers of the /Phylloglandula and the /Megistohibiscus clades
TABLE 2. Branch-based phylogenetic definitions for clades referred to
in this paper
/Euhibiscus Internal: H. rosa-sinensis L.
External: Urena lobata L., Hibiscus tiliaceus L., H.
trionum L., Kosteletzkya virginica (L.) C. Presl ex
A. Gray, and H. calyphyllus Cav.
/Megistohibiscus Internal: Megistostegium microphyllum Hochr. and
Hibiscus macrogonus Baill.
External: Alyogyne hakeifolia Alef., Hibiscus
surattensis L., H. rosa-sinensis L., H. dongolensis
Caill. ex Delile, and H. pentaphyllus F. Muell.
FIG. 1. Maximum likelihood trees from analysis of the individual data sets. A. ndhF;B.matK. The MP bootstrap support values (from 10,000 heuristic
searches) are given above the branches and Bayesian posterior probabilities are given below the branches. The placement of Helicteropsis, Hibiscus
grandidieri and Jumelleanthus (discussed in the text) are indicated with arrows.
2008] KOOPMAN & BAUM: MALAGASY HIBISCEAE 367
(indicated with TTCCC in Fig. 2), whereas GGGAA occurs in
all other sampled taxa (including Helicteropsis, Hibiscus gran-
dideri, and Jumelleanthus). Based on flat-weighted parsimony
on the optimal tree (matK alone or combined), we infer that
the GGGAA version is ancestral with two independent in-
versions, one on the stem lineage of /Megistohibiscus and
one on the stem lineage of /Phylloglandula. Such homoplasy
is plausible given the large number of inversions (4–200bp in
length) that have been recognized in plastid genes (reviewed
in Graham et al. 2000) and the potential of stem loop forma-
tion to facilitate the inversion process (Kelchner and Wendel
1996; Graham and Olmstead 2000).
Topology Tests—Templeton tests were implemented to
test certain phylogenetic and biogeographical hypotheses.
Because of the taxonomic history and interest in the bioge-
ography of Macrostelia we investigated the possibility that the
Malagasy accessions of Macrostelia and the Australian Hibis-
cus tozerensis (previously Macrostelia grandifolia) could form a
clade. The shortest tree compatible with such a constraint is
significantly longer than the unconstrained tree (cost = 15
steps; p= 0.0026). Likewise Templeton tests significantly re-
ject the hypothesis that there is a clade that includes all Mala-
gasy Hibisceae (cost = 46 steps; p< 0.0001). Thus, at least two
migrations to Madagascar (or many migrations from Mada-
gascar) are needed to explain these sequence data. Further-
more, the data reject (p < 0.01) a clade composed only of
Malagasy endemics sampled from /Euhibiscus (Macrostelia,
H. caerulescens, and H. humbertianus), as well as a clade com-
posed of all /Euhibiscus taxa found on Madagascar (Macro-
stelia,H. caerulescens,H. humbertianus, and H. ferrugineus)(p<
0.01). The lattermost taxon, H. ferrugineus, also occurs in Af-
rica, but deletion of this taxon did not alter the significant
rejection. Likewise, analysis of the matK data, including H.
grandidieri,Jumelleanthus, and Helicteropsis, significantly reject
(p < 0.01) a single, exclusively Malagasy subclade. Thus, the
/Euhibiscus clade seems either to have invaded Madagascar
on multiple occasions and/or there have been migrations
from Madagascar to Asia/Australasia.
Molecular Dating—Molecular dating methods approxi-
mate the crown node of Hibisceae at 15–19 Ma (Table 3; Fig.
4). /Megistohibiscus is estimated to have colonized Mada-
gascar sometime between ca. 11.5 Ma (crown age) and 14.5
Ma (stem age). The second invasion of Madagascar in the
common ancestor of H. ferrugineus and Macrostelia is esti-
mated to have occurred ca. 3.7–4.9 Ma. These dates are, how-
ever, associated with broad confidence intervals (Table 3).
FIG. 2. Flat-weighted parsimony reconstruction of structural molecu-
lar characters on one of the 12 MP trees from the combined matK and ndhF
data. Solid squares with lowercase letters indicate indels in the matK
sequences, hollow squares with uppercase letters correspond to events in
the ndhF sequences. Letters above squares identify certain indels, num-
bers below indicate the number of bases involved in the event (- = dele-
tion, + = insertion). Clades marked with TTCCC and an inverted arrow
highlight the inversion discussed in the text. Clade names are indicated to
FIG. 3. Maximum likelihood tree from the combined matK and ndhF
data. The MP bootstrap support values (from 10,000 heuristic searches)
are given above the branches and Bayesian posterior probabilities are
given below the branches. Internal branches marked with arrows are not
judged significantly greater than zero length using a likelihood ratio test.
368 SYSTEMATIC BOTANY [Volume 33
Phylogenetic Relationships of Malagasy Species Tradi-
tionally Placed in Hibiscus—We included a Malagasy acces-
sion of the widespread mangrove species Hibiscus tiliaceus
sect. Azanza and found that it was placed, as expected, in the
/Phylloglandula clade (/Azanza subclade). Likewise, there
is little reason to doubt that Malagasy accessions of non-
Malagasy endemic species in Hibiscus sects. Furcaria,Trio-
num, and Ketmia, will affiliate with their relatives in their
respective clades. However, all sampled Malagasy endemic
species of Hibiscus were found in one of two clades: /Euhi-
biscus and /Megistohibiscus.
The combined data showed that Malagasy endemic species
traditionally assigned to H. sects. Bombicella, Ketmia, Tricho-
spermum, and Spatula were members of the /Euhibiscus
clade. Unpublished molecular data supports the inclusion of
all other Malagasy endemic members of Hibiscus sects. Bom-
bicella and Ketmia to /Euhibiscus. Though no Malagasy ac-
cessions from H. sects. Solandra and Lilibiscus (following Ho-
chreutiner 1955) were sampled in this study we suspect that
they will probably be added to /Euhibiscus based on the
presence of free epicalyx lobes and non-inflated calices—
traits they share with H. sects. Bombicella, Ketmia, Trichosper-
mum and Spatula.MatK data and epicalyx morphology sup-
ported Hibiscus grandidieri (H. sect. Trichospermum) and the
monotypic endemic genera Jumelleanthus and Helicteropsis as
members of this clade as well. Although the latter result is
contradicted by ndhF, we suspect that this is due to anoma-
lous molecular evolution of this gene. Additionally, the en-
demic genus Macrostelia is placed in the /Euhibiscus clade.
All Malagasy endemic species from Hibiscus sect. Azanza
were found to be members of /Megistohibiscus. Contradict-
ing traditional taxonomy, they are not closely related to
Asian taxa from H. sect. Azanza (or the widespread H.tili-
aceus). While /Megistohibiscus also includes Malagasy spe-
cies of Kosteletzkya and contains three endemic segregate gen-
era (Humbertiella, Megistostegium, and Perrierophytum), it does
not include exemplars sampled from any other traditionally
recognized sections of Hibiscus. It would be premature, how-
ever, to rule out the possibility that some species that Hoch-
reutiner (1955) assigned to sections other than Azanza (e.g.
Hibiscus lasiococcus in section Columnaris) could also turn out
to be in the /Megistohibiscus clade. Members of /Megisto-
hibiscus, in contrast to /Euhibiscus, have epicalyx lobes that
are fused to some extent to one another, but not to the calyx.
Based on this morphological trait, we expect that the remain-
ing 9 species in H. sect. Azanza on Madagascar, as well as all
17 unsampled species of Humbertiella, Megistostegium, Perri-
erophytum, and Malagasy Kosteletzkya will fall into the
The data presented here amplify previous concerns about
the nomenclature and taxonomy of Hibiscus. Pfeil and Crisp
(2005) proposed that Hibiscus be equated with the major
clade that includes all Hibisceae except the clade we here
name /Megistohibiscus. Thus, they proposed ultimately
transferring all species from genera such as Decaschistia,Fio-
ria,Pavonia,Kydia, and Abelmoschus into Hibiscus. Given their
proposed strategy, our data would appear to add Macrostelia,
Jumelleanthus, and Helicteropsis to the list of segregate genera
needing to be subsumed within Hibiscus s.l.
The more difficult taxonomic question is how to accom-
modate /Megistohibiscus in a ranked generic system, while
only recognizing monophyletic taxa. Three options present
themselves: (1) expanding Hibiscus to include /Megistohibis-
cus; (2) treating /Megistohibiscus at the generic rank (either
Megistostegium or Perrierophytum, which were published si-
multaneously, would be the correct name for this genus), or;
(3) continuing to recognize Humbertiella,Perrierophytum (per-
haps expanded to include some Kosteletzkya, see below), and
Megistostegium, and providing new generic names for other
monophyletic lineages within /Megistohibiscus. All of these
choices would require the publication of new combinations
and significant changes in the circumscription of tradition-
ally recognized genera. Options 1 and 2 (but not option 3)
would also entail the dissolution of well-demarcated and
Kosteletzkya—The genus Kosteletzkya was originally pro-
posed to accommodate New World members of Hibisceae
with 1-seeded mericarps (Presl 1831). It was later expanded
to include a number of Old World species and, as presently
circumscribed, has 17 species (Fryxell 1988), eight of which
are endemic to Madagascar and one, K. diplocrator, which
occurs in both Madagascar and continental Africa. The Mala-
gasy species have been assigned to two sections: K. section
Azanzoides with three species of robust shrubs or small trees
with globose capsules, and K. sect. Eukosteletzkya with five
subshrubs with angular fruit.
Pfeil and Crisp (2005) found that Kosteletzkya s.s. belongs in
the /Trionum clade. The placement of the North American K.
virginica in a clade with Pavonia,Malvaviscus, and Abel-
moschus in this study is consistent with that placement. How-
ever, all three Kosteletzkya from Madagascar sampled here
were found to be members of the /Megistohibiscus clade.
Schatz (2001) suggested that K. section Azanzoides was more
related to Hibiscus rather than to other Kosteletzkya. Consis-
tent with this general finding, we found that K. diplocrator is
TABLE 3. Divergence times for clades within /Eumalvoideae based on Penalized Likelihood analyses using different stratigraphic calibrations. Dates
in millions of years ago (Ma) for each node of interest for the original data generated in r8s (bold). Confidence intervals (96%) are given in parentheses
after dates. E. estalae =Echiperiporites estalae,M. bakonyensis =Malvacearumpollis bakonyensis. Numbers in parentheses after nodal names are node
assignments that correspond to nodes on the chronogram in Fig. 4.
E.estalae (40 Ma) E.estalae (37 Ma) M. bakonyensis (37 Ma) M. bakonyensis (35 Ma)
E.estalae (1) 40 37 46.6 (43.5–57.9) 44.7 (43.3–58.2)
M. bakonyensis (2) 32.1 (25.5–38.2) 29.7 (23.6–35.4) 37 35
Gossypieae (3) 15.8 (5.3–33.9) 14.6 (4.9–31.3) 17.3 (6.1–35.5) 16.4 (5.7–33.0)
Malveae (4) 12.2 (6.1–21.8) 11.3 (5.6–20.2) 14.3 (5.9–25.4) 13.4 (5.6–23.9)
Hibisceae (5) 16.7 (8.3–26.5) 15.5 (8.1–24.5) 19.3 (11.9–34.8) 18.1 (11.7–31.1)
/Euhibiscus (6) 9.9 (4.6–18.9) 9.2 (4.3–17.1) 11.6 (4.3–20.6) 10.9 (4.5–19.5)
Malagasy /Euhibiscus (7) 3.9 (0.4–10.2) 3.7 (1.3–9.4) 4.9 (1.7–10.1) 4.7 (1.7–9.5)
/Megistohibiscus (8) 12.5 (5.8–24.9) 11.45 (5.4–21.1) 14.5 (8.7–23.9) 13.7 (8.3–22.6)
2008] KOOPMAN & BAUM: MALAGASY HIBISCEAE 369
more closely related to species traditionally placed in Hibis-
cus sect. Azanza than to either Kosteletzkya s.s. or Malagasy K.
sect. Eukosteletzkya. Instead, the two sampled species of K.
sect. Eukosteletzkya appear to be closely related to the genus
Perrieriophytum with good support (BS = 99, PP = 100). A
relationship between Perrierophytum and K. sect. Eukosteletz-
kya is plausible given that these taxa share similar tomentose
leaves and distinctly bilobed corollas.
Macrostelia—The Malagasy Macrostelia species sampled
here, M. laurina, is more closely related to other Malagasy
Hibiscus than to the Australian Hibiscus tozerensis, with which
it has previously been allied based on penninerved leaf ve-
nation, calyx and epicalyx morphology, and a pubescent
style (Fryxell 1974a). Our unpublished matK data for Austra-
lian former Macrostelia (H. macilwraithensis (Fryxell) Craven &
B.E. Pfeil, and H. propulsator Craven & B.E. Pfeil) supports
FIG. 4. Calibrated chronogram based on a Penalized Likelihood analysis of the combined data, calibrating the basal node (E. estelae pollen) at 40 Ma.
Confidence intervals (96%) on nodal ages, estimated using non-parametric bootstrapping, are shown with thin black lines. The range of means across
the four calibrations used is shown using the thick grey bars. Numerical values of the 96% confidence intervals for ages of the numbered nodes can be
found in Table 3.
370 SYSTEMATIC BOTANY [Volume 33
their close relationship to H. tozerensis. In all analyses, these
Australian taxa are more closely related to H. syriacus (BS =
98) than to M. laurina.
While we have not sampled other Malagasy Macrostelia
species for molecular data, our observations of herbarium
specimens suggest that they form a cohesive morphological
group. Thus, we infer that M. laurina likely serves as a valid
exemplar for all species of Macrostelia. Considering that Mac-
rostelia is clearly embedded within /Euhibiscus, we presume
that it will ultimately be sunk into Hibiscus s.l. Nonetheless,
efforts should be made to collect Macrostelia species on Mada-
gascar’s east coast since the plants inhabit severely endan-
gered forest fragments and several species, some unde-
scribed, are already feared extinct.
Hibiscus humbertianus—This species was originally de-
scribed in Hibiscus but was transferred to Cienfuegosia (a ge-
nus in the tribe Gossypieae) because it possessed ‘gossypol
glands’(Fryxell 1974b). Recent tests for terpenoid gossypol
were negative and further investigations of gland morphol-
ogy in Hibiscus humbertianus did not match those of other
Cienfuegosia (Fryxell 1997). On this evidence, Fryxell sug-
gested that this taxon be returned to Hibiscus. The phyloge-
netic placement of H. humbertianus in this study (associated
with Malagasy members of /Euhibiscus) supports this deci-
Helicteropsis—Helicteropsis microsiphon (Baill.) Hochr. was
first named in Hibiscus (H. microsiphon Baill.), but was later
erected as a monotypic genus to recognize its small, tubular
flowers, exerted staminal column, and unusually shaped
leaves with a cordate base and emarginate apex (Ho-
chreutiner 1925b). Our data suggest a close relationship be-
tween Helicteropsis microsiphon and Hibiscus grandidieri H. Bn.
Both taxa are embedded in the /Euhibiscus clade with high
support based on the matK data (see Fig. 1B). These species
are small to large shrubs and grow on calcareous soils. Both
species have a reduced epicalyx (1–3–5mm), a partially to
fully tubular calyx, and a bright red corolla. The staminal
column and style are exerted well past the corolla in both
taxa: the style is twice the length of the perianth in H. gran-
didieri and 3–4.5 times the length of the perianth in Helicterop-
sis. Thus, quite apart from the prospect of wholesale move-
ment of segregate genera into Hibiscus, the phylogenetic data
render it more difficult to justify the recognition of the mo-
notypic genus Helicteropsis.
Jumelleanthus—This monotypic, endemic genus was es-
tablished by Hochreutiner (1924) to encompass shrubs in the
Sambirano Valley of northwestern Madagascar with large
foliaceous stipules, an involucre of three, free, foliaceous
bracteoles and two ovules per locule. Hochreutiner placed
the genus in Hibisceae, but Bayer and Kubitzki (2003) later
placed it incertae sedis. Although the taxonomic assignment of
Jumelleanthus remains uncertain due to discordance between
the matK and ndhF data, the epicalyx morphology is concor-
dant with the matK placement of J. perrieri in the /Euhibiscus
clade. Further analysis of Jumelleanthus should be a priority
for future systematic research in Malvoideae.
Perrierophytum—Hochreutiner (1915) erected the genera
Perrierophytum and Perrieranthus, but later synonymized the
two (Hochreutiner 1925a). The most recent treatment (Ho-
chreutiner 1955) recognizes nine species, although more spe-
cies remain undescribed (M. M. Koopman pers. obs.). The
genus is distributed throughout the western portions of the
island, extending to Juan de Nova Island in the Mozambique
Channel (P. glomeratum). The group manifests diverse floral
morphology and petal color, but is united by leaves with
pellucid dots, inflorescences in paniculate cymes, and an epi-
calyx that equals or surpasses the globular to tubular calyx,
which in turn conceals a reduced bilobed corolla. The two
species sampled here form a well-supported clade (BS = 95,
PP = 100) and are closely related to members of Kosteletzkya
sect. Eukosteletzkya. This affinity with Kosteletzkya was noted
early in the literature (Hochreutiner 1915) and molecular evi-
dence here confirms the close relationship between the
shrubby Malagasy endemic Kosteletzkya and Perrierophytum
(BS = 99, PP = 100).
Humbertiella—Humbertiella was erected by Hochreutiner
(1926) based on a single species with uncertain familial affili-
ation. When another species was added, Hochreutiner (1932)
decided that the genus belonged in Bombacaceae with the
interpretation that the sessile anthers were bilocular. He later
erected Neohumbertiella (Hochreutiner 1940) to accommodate
two ‘monothecal’species that he thought were more appro-
priately placed in Malvaceae. The two genera were subse-
quently synonymized (Hutchinson 1967), and the genus was
expanded and revised to include six shrubby species re-
stricted to southwest Madagascar (Dorr 1990). The genus is
unusual for its elaborated and inflated apical teeth (known as
coronules) and flattened capsules subtended by an accrescent
or erect calyx. The monophyly of this genus is well supported
here (BS = 100, PP = 100) and it is shown to be sister to the
genus Perrierophytum s.l. (BS = 77, PP = 0.99).
Megistostegium—Originally described as Hibiscus nodulo-
sus (Drake 1903), and later transferred in and out of an im-
proper use of the name Macrocalyx (Costantin and Poisson
1908; Poisson 1912), the genus Megistostegium was ultimately
erected in 1915 to accommodate three species restricted to the
Mahafaly Plateau and associated coastline in the most south-
western arid scrublands of Madagascar. The flowers of this
genus have a distinctive, red, 4-lobed campanulate epicalyx.
Given this morphological uniqueness, it is not surprising that
the two species sampled form a highly supported clade (BS =
100, PP = 100). These bird-pollinated (M. M. Koopman pers.
obs.) treelets and subshrubs are sister to the endemic genera
Humbertiella and Perrierophytum s.l., which are also centered
in the arid southwest.
Biogeographic History of Malvoideae—/Eumalvoideae
are widely distributed throughout the subtropics and tem-
perate regions of the world. Pfeil et al. (2002) hypothesized
that Malvoideae has an Australasian center of origin and that
the separation of Malvoideae from the predominantly New
World Bombacoideae corresponded to Gondwanan fragmen-
tation. Baum et al. (2004) noted that Malvoideae emerge from
a basal grade of Neotropical taxa (Matisieae, Pentaplaris, and
Uladendron) and proposed, instead, transPacific dispersal
from South America to Australia. This dispersalist model is
supported indirectly by the fact that some of these early
branching lineages (Camptostemon and Lagunaria) are well
adapted for oceanic dispersal, being mangrove or littoral
The phylogenetic resolution of the Malagasy /Megistohi-
biscus clade as sister to the remainder of Hibisceae is consis-
tent with the view that the Indian Ocean Basin served as the
cradle for the initial radiation of Malvoideae. Although taxon
sampling in Hibiscus s.l. remains incomplete, it is noteworthy
that several of the early branching lineages have circum-
Indian Ocean distribution, most notably the early divergent
2008] KOOPMAN & BAUM: MALAGASY HIBISCEAE 371
clades /Calyphylli (H. dongolensis; Africa, H. calyphyllus not
sampled here but see Pfeil et al. 2002; Australia), /Pentaphyl-
lus (not sampled here but see Pfeil et al. 2002; Africa),
/Azanza (Tropical Asia), and Kydieae (Kydia; India/Sri
Lanka). Furthermore, our molecular dating analyses, while
tentative, support a radiation of Hibisceae in the early to mid
Miocene (ca. 19.5–15.5 Ma), which is too late to be explained
by vicariance. Thus, frequent long-distance dispersal appears
to have played an important role in the radiation of Hibis-
ceae. Similar transoceanic dispersal has been suggested as a
major factor in the evolution of Gossypieae, in the form of
salt-water resistant seeds traveling across the ocean (DeJoode
and Wendel 1992; Wendel and Albert 1992), and in Adan-
sonia, in the form of buoyant fruit that are inferred to have
traversed the Indian Ocean, ca. 11 Ma (Baum et al. 1998).
While acknowledging the limits of our molecular dating
analyses, both due to uncertain placement of the fossil pollen
forms and broad confidence intervals based on the stochas-
ticity of molecular evolution, our study provides useful ini-
tial estimates of divergence times for different lineages of
Hibisceae. Our results are compatible with those of Seelanan
et al. (1997), who calibrated their study using an average rate
of nucleotide substitution for ndhF to be 5x10
per site per year (after Palmer 1991). They inferred that Gos-
sypioides and Kokia diverged from one another in the mid
Pliocene (3 Ma) and from Gossypium in the mid Miocene (12.5
Ma), whereas we date the Gossypioides/ Kokia divergence at
4.8 Ma and its divergence from Gossypium at 12 Ma.
Our best estimate of the arrival of /Megistohibiscus in
Madagascar is in the mid Miocene (11.5–14.5 Ma), whereas it
appears that the /Euhibiscus clade did not arrive until ap-
proximately 6–11 Ma later. The most parsimonious inference
based on our sampling is that there was one dispersal from
Madagascar to give rise to the predominantly Indo-Pacific
Hibiscus sect. Lilibiscus group (here represented by H. rosa-
sinensis and H. waimeae). However, given the sparse sampling
in this clade, we cannot rule out multiple independent dis-
persals to Madagascar. Further sampling in /Euhibiscus is an
obvious priority for future research.
The Diversification of Malvaceae in Madagascar—Our
data suggest that at least two and possibly more independent
lineages of Hibisceae colonized Madagascar and underwent
a significant autochthonous radiation. These radiations each
encompass the full geographic expanse of Madagascar, al-
though the /Megistohibiscus clade is absent from Eastern
rainforest and proportionately more abundant in the arid
South and West. Also, each radiation has yielded species of
sufficient morphological distinctiveness to have been af-
forded generic status: Helicteropsis, Jumelleanthus, and Macro-
stelia in /Euhibiscus and Humbertiella, Megistostegium, and
Perrierophytum in /Megistohibiscus. Thus, while the number
of species of Hibisceae on Madagascar is not especially great,
given the relative antiquity of its component lineages, Mada-
gascar stands out for the morphological diversity that it
manifests. In particular, the Humbertiella-Megistostegium-
Perrierophytum subclade is composed of species that are strik-
ingly divergent from conventional Hibiscus (Humbertiella was
formerly placed in Bombacaceae) and from one another. In-
deed these three traditional genera, comprising 18 described
species, seem to encompass almost as much variation in flo-
ral morphology as the rest of Hibisceae. Is there something
about Madagascar that can explain such dramatic examples
of diversification, for example, low competition, high eco-
logical heterogeneity, or edaphic restriction? The Malagasy
Hibisceae may prove to be a suitable group to further explore
this fascinating question.
ACKNOWLEDGMENTS. The authors acknowledge the National Science
Foundation grant DEB-0416096 for funding this work. We also thank the
Missouri Botanical Garden Madagascar Research and Conservation Pro-
gram for their logistical and administrative support. Invaluable field as-
sistance was provided by Roland Ranaivojaona, Hanta Razafindraibe,
and Jacqueline Razanatsoa. We thank Stacey Dewitt Smith and Rebecca
Oldham-Haltom for assistance with sequencing and Kandis Elliot for
assistance with artwork. Comments on the manuscript and the project in
general were kindly provided by Stewart Hinsely, Bernard Pfeil, Susanne
Renner, George Schatz, and Randy Small. The authors thank Kweon Heo,
Bernard Pfeil, Randy Small, Richard Razakamalala, Laurent Gautier, and
S. Wolhauser for sharing DNA and for help in obtaining leaf material of
Alverson, W. S., B. A. Whitlock, R. Nyffeler, C. Bayer, and D. A. Baum.
1999. Phylogeny of the core Malvales: evidence from ndhF sequence
data. American Journal of Botany 86: 1474–1486.
Baldwin, B. G. and M. J. Sanderson. 1998. Age and rate of diversification
of the Hawaiian silversword alliance (Compositae). Proceedings of the
National Academy of Sciences USA 95: 9402–9406.
Baum, D. A., R. L. Small, and J. F. Wendel. 1998. Biogeography and floral
evolution of Baobabs (Adansonia, Bombacaceae) as inferred from
multiple data sets. Systematic Biology 47: 181–207.
Baum, D. A., S. D. Smith, A. Yen, W. S. Alverson, R. Nyffeler, B. A.
Whitlock, and R. L. Oldham. 2004. Phylogenetic relationships of
Malvatheca (Bombacoideae and Malvoideae: Malvaceae sensu lato)
as inferred from plastid DNA sequences. American Journal of Botany
Bayer, C. and L. J. Dorr. 1999. A synopsis of the Neotropical genus Pen-
taplaris, with remarks on its systematic position within core Mal-
vales. Brittonia 51: 134–148.
Bayer, C. and K. Kubitzki. 2003. Malvaceae. Pp. 225–311 in The families and
genera of vascular plants vol. 5, eds. K. Kubitzki and C. Bayer. Berlin:
Christensen, P. B. 1986. Evolutionary trends in the pollen morphology in
Malvaceae. Pp. 425–427 in Pollen and spores: form and function, eds. S.
Blackmore and I. K. Ferguson. London: Academic Press.
Costantin, J. N. and H. L. Poisson. 1908. Macrocalyx tomentosa.Comptes
Rendus des Séances de l’Academie des Sciences Paris 147: 637.
Craven, L. A. and B. E. Pfeil. 2004. Australian representatives of Macro-
stelia transferred to Hibiscus (Malvaceae), with the description of a
new species. Adansonia 26: 235–240.
DeJoode, D. R. and J. F. Wendel. 1992. Genetic diversity and origin of the
Hawaiian islands cotton, Gossypium tomentosum. American Journal of
Botany 79: 1311–1319.
Dorr, L. J. 1990. An expansion and revision of the Malagasy genus Hum-
bertiella (Malvaceae). Bulletin du Muséum d’Histoire Naturelle 4e sér.
sect B, Adansonsia 12: 7–27.
Drake del Castillo, E. 1903. Hibiscus nodulosus sp. nov. Bulletin du Muséum
d’Histoire Naturelle 9: 38.
el Naggar, S. M. 2004. Pollen morphology of Egyptian Malvaceae: an
assessment of taxonomic value. Turkish Journal of Botany 28: 227–240.
Farris, J. S., M. Kallersjo, A. G. Kluge, and C. Bult. 1994. Testing signifi-
cance of incongruence. Cladistics 10: 315–319.
Fryxell, P. A. 1974a. New species of Gossypium, Decaschistia, and Macro-
stelia (Malvaceae) from Australia. Australian Journal of Botany 22: 183–
Fryxell, P. A. 1974b. Cienfuegosia Cav. extended to Madagascar. Annals of
the Missouri Botanical Gardens 61: 491–493.
Fryxell, P. A. 1988. Malvaceae of Mexico. Systematic Botany Monographs 25:
Fryxell, P. A. 1997. The American genera of Malvaceae II. Brittonia 49:
Germeraad, J. H., C. A. Hopping, and J. Muller. 1968. Palynology of
Tertiary sediments of tropical areas. Review of Palaeobotany and Pal-
ynology 6: 189–348.
Graham, S. W., P. A. Reeves, A. C. E. Burns, and R. G. Olmstead. 2000.
Microstructural changes in noncoding chloroplast DNA: Interpreta-
tion, evolution, and utility of indels and inversions in basal angio-
372 SYSTEMATIC BOTANY [Volume 33
sperm phylogenetic inference. International Journal of Plant Sciences
Graham, S. W. and R. G. Olmstead. 2000. Evolutionary significance of an
unusual chloroplast DNA inversion found in two basal angiosperm
lineages. Current Genetics 37: 183–188.
Hochreutiner, B. P. G. 1915. Malvacées de Madagascar: Perrierophytum,
Perrieranthus et Megistostegium. Annuaire du Conseratoire et du Jardin
Botanique de Genève 18: 215–237.
Hochreutiner, B. P. G. 1924. Genres nouveaux et genres discutésdela
famille des Malvacées; Jumelleanthus. Candollea 2: 79–83.
Hochreutiner, B. P. G. 1925a. Réforme et extension du genre Perrierophy-
tum Hochr. Candollea 2: 145–154.
Hochreutiner, B. P. G. 1925b. Encore un genre nouveau de Malvacées de
Madagascar. Candollea 2: 155–158.
Hochreutiner, B. P. G. 1926. Humbertiella. Un genre nouveau de Mal-
vacées de Madagascar. Candollea 3: 1–4.
Hochreutiner, B. P. G. 1932. Extension et affinities du genre Humbertiella
Hochr. Candollea 5: 1–4.
Hochreutiner, B. P. G. 1940. Neohumbertiella. Nouveau genre de Mal-
vacées. Candollea 8: 27–34.
Hochreutiner, B. P. G. 1952. Macrostelia, un nouveau genre extraordinaire
de Malvacées de Madagascar. Notulae Systematicae 14: 229–234.
Hochreutiner, B. P. G. 1955. Malvacées. Pp. 1–170 in Flore de Madagascar et
des Comores vol. 129, ed. H. Humbert. Paris: Firmin-Didot.
Huelsenbeck, J. P. and F. Ronquist. 2001. MrBayes: Bayesian inference of
phylogenetic trees. Bioinformatics (Oxford, England) 17: 754–755.
Hutchinson, J. 1967. The genera of flowering plants (Angiospermae). Dicoty-
ledones. Vol. 2. Oxford: Clarendon Press.
Kelchner, S. A. and J. F. Wendel. 1996. Hairpins create minute inversions
in non-coding regions of chloroplast DNA. Current Genetics 30: 259–
Lewis, P. O., M. T. Holder, and K. E. Holsinger. 2005. Polytomies and
Bayesian phylogenetic inference. Systematic Biology 54: 241–253.
MacPhail, M. K. 1999. Palynostratigraphy of the Murray Basin, inland
southeastern Australia. Palynology 23: 197–240.
MacPhail, M. K. and E. M. Truswell. 1989. Palynostratigraphy of the
central west Murray Basin. BMR Journal of Australian Geology and
Geophysics 11: 301–331.
Maddison, D. R. and W. P. Maddison. 2003. MacClade 4. Release version
4.06. Sunderland: Sinauer Associates.
Marcano-Berti, L. 1971. Uladendron, nuevo genero de las Malvaceae. Pit-
tieria 3: 9–17.
Muller, J. 1981. Fossil pollen records of extant angiosperms. Botanical
Review 47: 1–142.
Nilsson, S. and A. Robyns. 1986. Bombacaceae Kunth. Pp. 1–59 in World
Pollen and Spore Flora vol.14, eds. Almquist and Wiksell. Stockholm:
Nyffeler, R., C. Bayer, W. S. Alverson, A. Yen, B. A. Whitlock, M. Chase,
and D. A. Baum. 2005. Phylogenetic analysis of the Malvadendrina
clade (Malvaceae s.l.) based on plastid DNA sequences. Organisms,
Diversity & Evolution 5: 109–123.
Nylander, J. A. A. 2004. MrModeltest v2. Program distributed by the
author. Evolutionary Biology Centre, Uppsala University.
Palmer, J. D. 1991. Plastid chromosome: structure and evolution. Pp. 5–53
in Cell culture and somatic cell genetics of plants vol. 7, eds. L. Bogorad
and I. K. Vasil. San Diego: Academic Press.
Paradis, E., J. Claude, and K. Strimmer. 2004. APE: Analyses of phyloge-
netics and evolution in R language. Bioinformatics (Oxford, England)
Paradis, E., K. Strimmer, J. Claude, G. Jobb, R. Opgen-Rhein, J. Dutheil, Y.
Noel, and B. Bolker. 2006. APE: Analyses of phylogenetics and evo-
lution. R package version 1.8–2.
Pfeil, B. E. and M. D. Crisp. 2005. What to do with Hibiscus? A proposed
nomenclatural resolution for a large and well known genus of Mal-
vaceae and comments on paraphyly. Australian Systematic Botany 18:
Pfeil, B. E., C. L. Brubaker, L. A. Craven, and M. D. Crisp. 2002. Phylog-
eny of Hibiscus and the tribe Hibisceae (Malvaceae) using chloroplast
DNA sequences of ndhF and the rpl16 intron. Systematic Botany 27:
Poisson, H. L. 1912. Macrocalyx tomentosa Cost. et Poisson. Fl. Mérid. Mada-
gascar Série A 605: 26–27.
Presl, C. 1831. Kosteletzkya Presl. Reliquiae Haenkeanae 2. 130 t. 70.
R Development Core Team. 2005. R: a language and environment for
statistical computing. Vienna: R Foundation for Statistical Comput-
Rambaut, A. and N. C. Grassly. 1997. Seq-Gen: an application for the
Monte Carlo simulation of DNA sequence evolution along phyloge-
netic trees. Computer Applications in the Biosciences 13: 235–238.
Sanderson, M. J. 2002. Estimating absolute rates of molecular evolution
and divergence times: A penalized likelihood approach. Molecular
Biology and Evolution 19: 101–109.
Sanderson, M. J. 2003. r8s: inferring absolute rates of molecular evolution
and divergence times in the absence of a molecular clock. Bioinfo-
matics 19: 301–302.
Schatz, G. E. 2001. Generic Tree Flora of Madagascar. Kew: Royal Botanical
Seelanan, T., A. Schnabel, and J. F. Wendel. 1997. Congruence and con-
sensus in the cotton tribe (Malvaceae). Systematic Botany 22: 259–290.
Swofford, D. L. 2001. PAUP*b10: Phylogenetic analysis using parsimony.
Sunderland: Sinauer Associates.
Takayama, K., T. Ohi-toma, H. Kudoh, and H. Kato. 2005. Origin and
diversification of Hibiscus glaber, species endemic to the oceanic Bo-
nin Islands, revealed by chloroplast DNA polymorphism. Molecular
Ecology 14: 1059–1071.
Tate, J. A., J. F. Aguilar, S. J. Wagstaff, J. C. La Duke, T. A. Bodo Slotta, and
B. B. Simpson. 2005. Phylogenetic relationships within the tribe
Malveae (Malvaceae, subfamily Malvoideae) as inferred from ITS
sequence data. American Journal of Botany 92: 584–602.
Templeton, A. R. 1983. Phylogenetic inference from restriction endonucle-
ase cleavage site maps with particular reference to the evolution of
humans and the apes. Evolution 37: 221–244.
Wendel, J. F. and V. A. Albert. 1992. Phylogenetics of the cotton genus
(Gossypium): character-state weighted parsimony analysis of chloro-
plast-DNA restriction site data and its systematic and biogeographic
implications. Systematic Botany 17: 115–143.
Wilmé, L., S. M. Goodman, and J. U. Ganzhorn. 2006. Biogeographic
evolution of Madagascar’s microendemic biota. Science 312: 1063–
Wood, G. R. 1986. Late Oligocene to Early Miocene palynomorphs from
GSQ Sandy Cape 1–3R. Geological Survey of Queensland, Publication
APPENDIX 1. Taxa, Genbank accession numbers (ndhF, matK; —=se-
quence not obtained) and vouchers of plant material from which DNA
was extracted for sequencing. Sequences generated in previous studies
are referenced with Genbank accession numbers (ndhF, matK; —=se-
quence not obtained). Taxa are listed alphabetically by genus and species.
Abelmoschus.Abelmoschus manihot (L.) Medik.; B. Pfeil et al. (2002)
AF384639, USA R. Oldham s.n. UW GH #408(WIS) EF562457. Abutilon.
Abutilon hybridum Hort.; Alverson et al. (1999); AF111716, Baum et al.
(2004); AY589058. Alyogyne. Alyogyne hakeifolia Alef.; B. Pfeil et al. (2002)
AF384658, Baum et al. (2004) AY589059. Camptostemon.Camptostemon
schultzii Mast.; Alverson et al. (1999) AF111727, Nyffeler et al. (2005)
AY321162. Decaschistia.Decaschistia byrnesii Fryxell; Baum et al. (2004)
AY589079, AY589066. Gossypioides.Gossypioides kirkii (Mast.) Skovsted;
Seelanan et al. (1997) U55329, AF403563. Gossypium. Gossypium hirsutum
L.; Seelanan et al. (1997) U55340, Nyffeler et al. (2005) AY321158. He-
licteropsis.Helicteropsis microsiphon (Baill.) Hochr.; Madagascar, P. B.
Phillipson 1955 (TAN); EF207296, EF207264. Hibiscus. Hibiscus bojerianus
Baill.; Madagascar, D. Baum 383 (MO); EF207306, EF207275. Hibiscus cae-
rulescens Baill.; Madagascar, M. Koopman 233 (MO); EF207297, EF207265.
Hibiscus cannabinus L.; USA, R. Small s.n. (TENN) seed obtained from
USDA GRIF1541; EF207290, EF207259. Hibiscus costatus A. Rich.; Seelanan
et al. (1997) U55323, Baum et al. (2004) AY589057. Hibiscus dongolensis
Caill. ex Delile; USA, R. Small s.n. seed obtained from USDA PI364899;
EF207303, EF207271. Hibiscus ferrugineus Cav.; Madagascar, M. Koopman
235 (MO); EF207300, EF207268. Hibiscus grandidieri Baill.; Madagascar, R.
Ramanjananhary 182 (TAN); EF207295, EF207263. Hibiscus hamabo Siebold
& Zucc.; China, S. W. Lee s.n. (Yunan University); EF207292, Takayama et
al. (2005), AB181099. Hibiscus humbertianus Hochr.; Madagascar, DuPuy,
DuPuy R. & Rovonjiansoa s.n. (P); EF207298, EF207266. Hibiscus macrogo-
nus Baill.; Madagascar, M. Koopman 289 (MO); EF562456, EF207273. Hi-
biscus macrophyllus Roxb.; Takayama et al. (2005) AB181076, AB181100.
Hibiscus mandrarensis Humbert ex Hochr.; Madagascar, P. B. Philipson et
al. 3978 (TAN); EF207305, EF207274. Hibiscus rosa-sinensis L.; Baum et al.
(2004) AY589075, Nyffeler et al. (2005) AY321160. Hibiscus surattensis L.;
USA, R. Small s.n. (TENN) seed obtained from USDA PI405511; EF207289,
EF207258. Hibiscus syriacus L.; USA, R. Small s.n. cultivated (TENN);
EF207302, EF207270. Hibiscus tiliaceus L.; Takayama et al. (2005)
AB181074, AB181098. Hibiscus tozerensis Craven & B.E. Pfeil; Australia,
2008] KOOPMAN & BAUM: MALAGASY HIBISCEAE 373
extracted by B. Pfeil Garraway Ck, Australia; EF207301, EF207269. Hibis-
cus waimeae A.Heller; USA, M. Koopman s.n. WIS Greenhouse collected
#880903; EF207294, EF207262. Howittia. Howittia trilocularis F. Muell.;
Baum et al. (2004) AY589085, AY589065. Humbertiella. Humbertiella de-
caryi (Hochr.) L. J. Dorr; Madagascar, D. Baum 385 (MO); EF207310,
EF207279. Humbertiella henrici Hochr.; Madagascar, M. Koopman 236
(MO); EF207312, EF207281. Humbertiella quararibeoides Hochr.; Madagas-
car, D. Baum 389 (WIS); EF207311, EF207280. Jumelleanthus. Jumelleanthus
perrieri Hochr.; Madagascar, Wolhauser and Andriamalaza 432 (TAN);
EF207304, EF207272. Kokia. Kokia drynarioides Lewton; Seelanan et al.
(1997) U55330, AF40356. Kosteletzkya. Kosteletzkya diplocrater Hochr.;
Madagascar, D. Baum 387 (MO); EF207307, EF207276. Kosteletzkya reflexi-
flora Hochr.; Madagascar, M. Koopman 258 (MO); EF207314, EF207283.
Kosteletzkya velutina Garcke.; Madagascar, M. Koopman 215 (MO);
EF207313, EF207282. Kosteletzkya virginica (L.) Presl ex A.Gray; USA, R.
Small 230 (TENN); EF207288, EF207257. Kydia. Kydia calycina Roxb.; In-
dia, R. Neupaney s.n. (KWNU); EF207293, EF207261. Lagunaria. Lagunaria
patersonia (Andrews) G. Don; Baum et al. (2004) AY589084, AY589064.
Lavatera. Lavatera acerifolia Cav.; Nyffeler et al. (2005) AY326475,
AY321159. Macrostelia. Macrostelia laurina (Baill.) Hochr. & Humbert;
Madagascar, S. Malcomber 2806 (MO); EF207299, EF207267. Malope. Ma-
lope trifida Cav.; Baum et al. (2004); AY589076, AY589060. Malvaviscus.
Malvaviscus arboreus Cav.; Alverson et al. (1999); AF111718, Baum et al.
(2004) AY589061. Megistostegium. Megistostegium perrieri Hochr.; Mada-
gascar, R. Ranairojaona 534 (TAN); EF207308, EF207277. Megistostegium
microphyllum Hochr.; Madagascar, D. Baum 382 (MO); EF207309,
EF207278. Perrierophytum. Perrierophytum humbertii Hochr.; Madagascar,
P. B. Phillipson et al. 3480 (US); EF207315, EF207284. Perrierophytum rubrum
Hochr.; Madagascar, P. B. Phillipson 2543 (MO); EF207316, EF207285. Mo-
diola. Modiola caroliniana (L.) G. Don; USA, J. Beck 6298 (TENN);
EF207287, EF207256. Pavonia. Pavonia strictiflora (Hook.) Fryxell; —,
Baum et al. (2004) AY589056. Pavonia multiflora A. St. Hil.; Alverson et al.
(1999); AF111719, —.Radyera. Radyera farragei (F. Muell.) Fryxell &
Hashmi; Baum et al. (2004) AY589078, AY589063. Sphaeralcea. Sphaeralcea
angustifolia G. Don; USA, R. Small 306 (TENN); EF207286, EF207255.
Thespesia. Thespesia thespesioides (R. Brown ex Benth.) Fryxell; Seelanan et
al. (1997) U55326, Nyffeler et al. (2005); AY321161. Uladendron. Uladen-
dron codesuri Marcano-Berti; Baum et al. (2004); AY589080, AY589067.
Urena. Urena lobata L.; US, J. Beck 5143 (TENN); EF207291, EF207260.
374 SYSTEMATIC BOTANY [Volume 33